The properties of a cpYFP [circularly permuted YFP (yellow fluorescent protein)] reported to act as a superoxide sensor have been re-examined in Arabidopsis mitochondria. We have found that the probe has high pH sensitivity and that dynamics in the cpYFP signal disappeared when the matrix pH was clamped by nigericin. In contrast, genetic and pharmacological manipulation of matrix superoxide had no detectable effect on the cpYFP signal. These findings question the existence of superoxide flashes in mitochondria.
The development of genetically encoded probes that allow quantification of ROS (reactive oxygen species) at a cell-specific and subcellular resolution in living tissues, and in real-time, have facilitated a more sophisticated understanding of the role of ROS in signalling and development [1,2]. Measuring superoxide dynamics in living intact cells, i.e. in vivo, is of particular interest as in most systems superoxide is the precursor for most other ROS and RNS (reactive nitrogen species) such as hydrogen peroxide, the hydroxyl radical or peroxynitrite. In mitochondria, where superoxide is generated by the ETC (electron transport chain), superoxide links bioenergetic function and radical dynamics . The superoxide production rate is strongly determined by the functional state of the ETC and it can rise dramatically under stress . Mitochondrial superoxide dynamics are therefore key for the understanding of the interplay between cellular energetic state, redox signalling and stress responses. However, most of our understanding about mitochondrial superoxide has been gathered under in vitro conditions, i.e. from isolated organelles. A genetically encoded superoxide sensor therefore has the potential to provide physiologically meaningful insights into mitochondrial superoxide dynamics in a cellular context. A superoxide sensor based on a cpYFP [circularly permuted YFP (yellow fluorescent protein)] has been reported and provided evidence for ‘flashes’ of superoxide production in single mitochondria of cardiomyocytes [5,6]. This raised important questions about the dynamic nature of mitochondrial free radical processes in vivo.
However, doubts have been raised about the specificity of cpYFP as a reporter of superoxide in the mitochondrial matrix . In the present study we have characterized the dynamic fluorescence properties of cpYFP targeted to the matrix of plant mitochondria. Properties of the probe were validated in living plant cells and in isolated mitochondria, and the sensitivity of the probe to mitochondrial bioenergetics, superoxide production and pH were established.
MATERIALS AND METHODS
Plant lines and culture conditions
The mitochondrial-targeted cpYFP construct was engineered by combining the cpYFP coding region in-frame and downstream of the mitochondrial-targeting sequence from Nicotiana plumbaginifolia β-ATPase  under control of the CaMV (cauliflower mosaic virus) 35S promoter in the binary vector pH2GW7 , using Gateway technology (Invitrogen). Hairpin RNAi (RNA interference) constructs were composed of specific and unique fragments of the MnSOD [manganese SOD (superoxide dismutase); At3g10920] sequence, in sense and antisense orientation, in the pOpOff2 vector providing dexamethasone-inducible transcript knockdown . Agrobacterium-mediated transformation of Arabidopsis thaliana (accession Columbia, Col-0) was performed by ‘floral dip’ . Transformants were selected by hygromycin (pH2GW7) or kanamycin (pOpOff2) resistance. Knockdown efficiency of the MnSOD transcript was assessed by semi-quantitative RT (reverse transcriptase)–PCR. Homozygous T3 lines were selected by segregation of antibiotic resistance. For in vivo microscopy, seedlings were cultivated for 7–10 days on vertical half-strength Murashige and Skoog medium agar plates under long-day conditions. For mitochondrial isolation, seedlings were cultured in a hydroponic pot system for 14 days as described previously . For induction of RNAi, 10 μM dexamethasone was added to the medium. Immunoblotting of total seedling extract was performed using a polyclonal MnSOD-specific antiserum as described previously [13,14].
Mitochondrial isolation and assays
Mitochondria were isolated from 14-day-old Arabidopsis seedling cultures as described previously , except the pH of the wash buffers was adjusted with Tris base instead of KOH. Coupling was measured in a Clark-type oxygen electrode ‘Oxygraph’ (Hansatech Instruments) as described previously  in ‘basic incubation medium’ [0.3 M sucrose, 5 mM KH2PO4, 10 mM Tes, 10 mM NaCl, 2 mM MgSO4 and 0.1% fatty-acid-free BSA (pH 7.2 adjusted with KOH)].
Living roots were imaged in half-strength Murashige and Skoog medium. Staining with 200 nM MitoTracker Orange CMTMRos (Invitrogen) was performed as described previously . Isolated mitochondria were immobilized on to round glass cover slides by centrifugation (2000 g for 5 min at 4 °C) in 0.3 M sucrose and 10 mM Tes (pH 7.2 adjusted with Tris base). Immediately before imaging, mitochondria were energized by the addition of medium containing the required respiratory substrate. Microscopy was performed using a Zeiss LSM510 META confocal microscope (Carl Zeiss MicroImaging) equipped with laser lines for 488, 405 and 543 nm excitation. Imaging was carried out with a ×63 lens [Zeiss Plan-Apochromat ×63, NA (numerical aperture)=1.4, oil-immersion] in multi-track mode with line switching. cpYFP was excited at 488 and 405 nm, and MitoTracker Orange was excited at 543 nm. cpYFP or MitoTracker fluorescence were measured at 505–530 nm and 565–615 nm respectively. Laser output was kept under 2% of maximal power equalling <2.5 μW for any wavelength as measured in the defocused beam at the backplane of the objective. Pixel size was 70 nm and pixel dwell time was 1.3 μs. Time series were collected with time intervals between 2 and 5 s. Microscopy data were background-subtracted and analysed using custom software as described previously .
Mitochondria were suspended in ‘basic incubation medium’ in 96-well plates. The total assay volume was 200 μl. Fluorescence was measured at 25 °C in a Beckman Coulter DTX 880 Multimode Detector equipped with 485 and 405 nm excitation filters, and a 500–530 nm BP emission filter. Data were background-subtracted and processed in Microsoft Excel.
cpYFP flashes in active plant mitochondria
To study the dynamic behaviour of cpYFP in plant mitochondria, we created stable Arabidopsis transformants expressing cpYFP in the mitochondrial matrix. Mitochondrial localization of the cpYFP signal with excitation at either 488 nm or 405 nm was confirmed by co-localization with MitoTracker Orange (Supplementary Figure S1 at http://www.BiochemJ.org/bj/437/bj4370381add.htm) and by fluorescence of isolated purified mitochondria (Figure 1B). Probe behaviour was monitored in living root epidermal cells by CLSM (confocal laser-scanning microscopy). The cpYFP signal was relatively homogeneous between all mitochondria of a single cell, whereas the intensity between cells was somewhat variable. At the level of individual mitochondria, infrequent, spontaneous and transient increases in cpYFP fluorescence were observed with 488 nm excitation (Figure 1A). The cpYFP signal in a single mitochondrion increased transiently in intensity before returning to baseline fluorescence within seconds (Figures 1A and 1C, and Supplementary Movie S1 at http://www.BiochemJ.org/bj/437/bj4370381add.htm). Fluorescence with 405 nm excitation remained unchanged during the event (results not shown). The duration of each event was typically 10–20 s, and the degree of signal increase was 1.3–2.5-fold over baseline. Hence this phenomenon showed similar characteristics in time and amplitude as cpYFP ‘flashes’ in mitochondria of mammalian cells described by Wang et al. . We therefore adopt the term ‘flashes’ for the events we describe in the present paper in plant mitochondria.
CpYFP flashes in mitochondria of living root epidermis cells and in isolated mitochondria
Another similarity between our findings and those of Wang et al.  was the infrequent occurrence of flashes. On average approximately one flash could be observed when monitoring mitochondria in the cortical cytoplasm of a living root epidermal cell for 10 min. Low flash frequency, coupled with the motility of mitochondria in plant cells (moving into and out of the field of observation in three-dimensions), prevented continuous observation of single mitochondria and quantification of flash frequency. We therefore optimized a cell-free system using isolated mitochondria that allowed a more systematic assessment of flashes. Isolated mitochondria were immobilized on to a coverslip and monitored by confocal microscopy. Respiratory control ratios were >2.5 for succinate as the substrate in all experiments. Isolated respiring mitochondria also displayed cpYFP flashes with 488 but not 405 nm excitation (Figures 1B and 1D, and Supplementary Figure S2 and Supplementary Movie S2 at http://www.BiochemJ.org/bj/437/bj4370381add.htm). Spontaneous changes in cpYFP signal intensity were more frequent in this system, making flashes the norm rather than the exception at approximately 0.8 flashes/mitochondrion per min. Similar behaviour was observed with other respiratory substrates (malate/pyruvate and NADH). No flashes occurred in non-respiring mitochondria (state 1) or upon uncoupling (results not shown), suggesting that the presence of a pmf (protonmotive force) was a prerequisite for flashes to occur. Isolated mitochondria therefore provided an appropriate model system to evaluate further the reason for dynamic cpYFP behaviour in the mitochondrial matrix.
cpYFP responds dynamically to changes in mitochondrial respiratory state
cpYFP flashes in the mitochondrial matrix have been linked to respiratory activity and functional ETC . To understand the average probe behaviour in different bioenergetic states, we measured the population response of mitochondrial suspensions in a 96-well plate reader. Different respiratory states were established by substrate or inhibitor feeding. The cpYFP signal responded to those changes in a highly sensitive manner with 485 nm excitation (Figure 2). A supply of respiratory substrate led to a strong increase in probe signal. Substrates with different ETC entry points (succinate, complex II; pyruvate/malate, complexes I and II; and ferrocyanide, complex IV) triggered similar cpYFP responses, although ferrocyanide triggered a less pronounced signal increase. For consistency, succinate was used as a substrate in all subsequent measurements. The absolute signal varied between 3- and 6-fold depending on mitochondrial preparation and storage time of mitochondrial suspensions before use (probably correlating with degree of coupling), but the relative probe response was consistent between individual experiments. The switch from state 2 to state 3 caused by the addition of ADP resulted in a decrease in cpYFP fluorescence, followed by recovery when entering state 4 upon ADP depletion (Figure 2). Uncoupling by CCCP (carbonyl cyanide m-chlorophenylhydrazone) led to a decrease in signal close to the value of state 1. In contrast, fluorescence at 405 nm excitation remained largely stable over the course of the measurements (Supplementary Figure S3 at http://www.BiochemJ.org/bj/437/bj4370381add.htm). Expressing the data as a ratio of the 485/405 signal was used in a ratiometric approach, but was not required in the plate reader system. The overall pattern of cpYFP signal in different respiratory states appeared to correlate with the magnitude of pmf, and therefore suggests that cpYFP is responsive to a parameter linked to, or downstream of, pmf in plant mitochondria.
The response of cpYFP fluorescence to different respiratory substrates and states in isolated mitochondria
Manipulation of matrix superoxide levels does not have an impact on cpYFP dynamics
cpYFP has been used previously as a superoxide sensor based on a 4-fold increase in fluorescence in response to artificially generated superoxide in vitro . The correlation of cpYFP signal and pmf in isolated plant mitochondria is consistent with the expected changes in superoxide, as mitochondrial superoxide production increases as a function of pmf . However, other parameters linked to pmf could also be responsible for the altered cpYFP fluorescence intensity. To further evaluate superoxide sensitivity of cpYFP in plant mitochondria, the effect of different complex III inhibitors on the cpYFP signal was measured. Antimycin A inhibits complex III by binding to the ‘N’ site, strongly stimulating superoxide release, whereas myxothiazol binds to the ‘P’ site and does not increase superoxide production . We found that both inhibitors caused a decrease in the cpYFP signal (Figure 3A). This suggests that an inhibition of electron transport at complex III, rather than changes in superoxide levels, causes changes in cpYFP intensity. In addition, menadione, a redox-active quinone that mediates superoxide release in mitochondria , had no effect on the cpYFP signal (Supplementary Figure S4 at http://www.BiochemJ.org/bj/437/bj4370381add.htm).
CpYFP response to antimycin A, superoxide scavengers and lack of MnSOD in active isolated mitochondria
The superoxide sensitivity of cpYFP in plant mitochondria was further assessed by treatment of mitochondria under state 2 respiration with three different membrane-permeant superoxide scavengers. Treatments with tiron within the concentration range used by Wang et al.  had no effect on the cpYFP signal (Figure 3B). In addition tempol, an SOD mimetic that has been shown to alleviate ROS-induced mitochondrial abnormality in plants, when applied at similar concentrations, did not cause any effect . By contrast, MnTMPyP [manganese(III) tetrakis(1-methyl-4-pyridyl)porphyrin], another SOD mimetic, led to a significant decrease in fluorescence. However a proportional decrease was also observed at 405 nm excitation upon MnTMPyP treatment (Supplementary Figures S5A and S5B at http://www.BiochemJ.org/bj/437/bj4370381add.htm), and MnTMPyP also decreased baseline cpYFP fluorescence, decreasing the signal beyond its state 1 value (Supplementary Figure S5C). These observations suggest quenching of cpYFP fluorescence by MnTMPyP, which may be a result of the intense brown colour of the compound, rather than an effect due to superoxide removal.
Finally, the response of cpYFP in isolated mitochondria lacking matrix MnSOD was tested. For that purpose cpYFP was expressed in the mitochondrial matrix of a line containing an inducible RNAi construct for suppression of mitochondrial SOD expression. RNAi-mediated suppression of MnSOD was confirmed by immunoblot analysis. MnSOD protein was undetectable from day 4 after induction (Figure 3C, inset). Respiring mitochondria isolated 5 days after induction showed no significant increase in cpYFP signal compared with control (Figure 3C).
In conclusion, there was no significant experimental correlation between mitochondrial ROS and cpYFP fluorescence in isolated Arabidopsis mitochondria.
A high pKa makes cpYFP highly pH-sensitive in the mitochondrial matrix
As no direct evidence for any significant sensitivity of cpYFP to superoxide in plant mitochondria could be found, other parameters linked to pmf were assessed for their ability to modulate cpYFP fluorescence. Strong pH-dependence of cpYFP was reported previously by Wang et al.  and cpYFP fluorescence intensity in isolated Arabidopsis mitochondria was observed to increase with increasing pH of the medium (Supplementary Figure S6 at http://www.BiochemJ.org/bj/437/bj4370381add.htm). Re-evaluation of the pH properties of cpYFP in mitochondrial lysates revealed a pKa of approximately 8.6 with a 485 nm excitation (Figure 4A). In addition, the dynamic range of the probe was broad between pH 7 and 10. A >20-fold increase in fluorescence was observed upon alkalization. Addition of 5 μM CCCP did not lead to any changes in lysates (results not shown), confirming that cpYFP was well equilibrated with the medium. In contrast with excitation with 485 nm, the probe showed only minor pH sensitivity with 405 nm excitation (Figure 4B). This allows ratiometric expression of pH sensitivity, independently of probe concentration and without a significant effect on pKa or the dynamic range (Figure 4C). The strong pH-dependence of cpYFP raises the possibility that pH changes rather than superoxide may account for the observed probe dynamics in mitochondria.
pH-dependence of mitochondrial cpYFP fluorescence
cpYFP dynamics in respiring plant mitochondria represent pH changes
To assess whether pH changes are sufficient to account for the observed probe dynamics in respiring mitochondria, ΔpH was separated from pmf (and thereby superoxide production) using nigericin. pH was clamped by nigericin-mediated K+/H+ antiport, whereas the voltage component of pmf was maintained. Addition of nigericin in the presence of K+ ions not only decreased the cpYFP signal to the same degree as the uncoupler CCCP (Figure 5A), but it also abolished the increase in cpYFP signal upon energization from state 1 to state 2 (Supplementary Figure 5C). Clamping matrix pH was therefore sufficient to stop mitochondrial cpYFP dynamics in mitochondrial suspensions, independently of the presence of a pmf and associated superoxide release. Nigericin also completely abolished cpYFP flashes in single isolated mitochondria monitored by CLSM (Figures 5B and 5C, and Supplementary Movies S3 and S4 at http://www.BiochemJ.org/bj/437/bj4370381add.htm). Flashes were retained when the identical treatments were applied in K+-free medium (results not shown), demonstrating the specificity of pH clamping in this system. Overall, changes in matrix pH provided a sufficient explanation for the observed cpYFP signal dynamics in mitochondrial suspensions, as well as in single mitochondria.
The effect of nigericin on cpYFP fluorescence dynamics in isolated mitochondria
On the basis of the observation of cpYFP flashes in single plant mitochondria, we aimed to elucidate the basis of the probe response. Similar flashing events have been observed in single mitochondria of cardiomyocytes using cpYFP and interpreted as response superoxide bursts in the matrix based on superoxide sensitivity of the sensor in an in vitro assay . Using isolated mitochondria as a minimal model system allowed us to validate cpYFP properties in situ, using a reductionist approach. Isolated plant mitochondria lack most respiratory metabolites and co-factors allowing tight control of potential probe interactors . The results of the present study confirm previous observations that the mitochondrial cpYFP signal is linked to mitochondrial electron transport and metabolic status [5,6]. Wang et al.  showed that flash frequency was strongly inhibited when electron transport was dysfunctional or inhibited in cardiomyocytes. In our hands absolute signal intensity strongly depended on respiratory state, decreased upon ETC inhibition at complex III and flashes were only observed in energized mitochondria. However, we were unable to demonstrate a significant response to matrix superoxide; instead the dynamic behaviour of cpYFP was fully accounted for by its pronounced pH sensitivity.
pH sensitivity of cpYFP was reported by Wang et al. , but without any determination of pKa and the dynamic range. pH sensitivity is common in GFP (green fluorescent protein)-derived probes; however, it can often be ignored when the pKa of the probe is sufficiently distant from the assay pH. Wang et al.  ruled out transient changes in matrix pH being responsible for cpYFP flashes, based on the absence of flashes using a mitochondrial-targeted EYFP (enhanced YFP) as a pH sensor. However, the pKa of EYFP is approximately 7.1 , meaning that it will not be particularly responsive to increases in pH from the physiological value of 7.8–7.9 that has been determined in respiring animal and plant mitochondria in vivo [20,21]. Comparably high matrix pH is expected in isolated mitochondria under state 2 and 4 respiration, and a medium pH of 7.2. On the basis of the cpYFP signal, a switch from state 1 to 2 increases matrix pH by approximately 1 unit, i.e. to approximately 8.2, under the conditions used in the present study (see Supplementary Figure S6). The high pKa of cpYFP at approximately pH 8.6 means that cpYFP will be much more sensitive than EYFP to pH changes in the mitochondrial matrix. For example, given a resting pH of 7.9, a pH increase of 0.4 unit results in a ∼2-fold increase in cpYFP fluorescence. This is in contrast with EYFP that would only show a very subtle fluorescence change (∼1.1-fold) upon the same pH increase . Moreover, the observed probe responses to changes in respiratory states of isolated mitochondria at buffer pH 7.2 appears, at least qualitatively, in line with the expected pH changes occurring in the matrix. cpYFP is therefore a highly responsive pH sensor in the physiological context of the mitochondrial matrix.
Wang et al.  provided good evidence that cpYFP is sensitive to superoxide in vitro. It therefore remains possible that cpYFP can report mitochondrial superoxide in vivo. However, for reasons stated above, our view is that changes in cpYFP fluorescence in the mitochondrial matrix in plants are more likely to be related to changes in matrix pH in vivo. A mechanism for the proposed superoxide sensitivity of cpYFP would help to clarify the situation; indeed Wang et al.  have suggested the involvement of two cysteine residues. However, the reactivity of thiols with superoxide is low, which makes such a mechanism unlikely . In contrast, pH-dependence of cpYFP fluorescence provides a straightforward explanation for the observed probe responses: the chromophore can be protonated, inducing a reversible fluorescence increase .
The superoxide sensitivity of cpYFP has been scrutinized in a review by Muller . A major criticism included the decreased flash rate upon antimycin A treatment, which generally induces superoxide release into the matrix. Our treatments of isolated plant mitochondria with antimycin A also gave a decrease in cpYFP intensity rather than the inverse, as expected from elevated superoxide production. cpYFP fluorescence was low when respiration was not active, either due to a lack of substrate or inhibition of complex III, and no cpYFP flashes occurred in the absence of pmf. Although plant and mammalian mitochondria differ in many aspects, changes in pH could in principle resolve the apparent contradictions that have been pointed out by Muller . Recently Pouvreau  also detected flashes in mitochondria of mouse skeletal muscle cells using a ratiometric pericam sensor that contains the cpYFP sequence. MitoSOX, a superoxide-sensitive chemical dye, was then used as an in vivo control for superoxide sensitivity of the sensor. An increase in MitoSOX intensity for a mitochondrion undergoing a flash was suggested. Therefore it is possible that cpYFP flashes in different systems represent different physiological phenomena. However, as the MitoSOX signal only increases slightly during a flash, even though MitoSOX acts as an irreversible and cumulative superoxide stain (with its specificity in vivo having been questioned ), the possibility of photo-oxidative ROS production during a transient increase in cpYFP fluorescence should be considered as an alternative explanation.
We suggest that cpYFP is a promiscuous sensor that can, in principle, respond to different parameters via changes in signal intensity. Its responsiveness to superoxide and thiol agents in vitro has been documented along with its pH sensitivity, and additional effectors cannot be ruled out. However, its pH response is substantial particularly at mitochondrial pH values, and cpYFP dynamics in mitochondria can be abolished when pH is clamped. These findings cast doubt on the existence of superoxide flashes in mitochondria. Moreover they emphasize the vital need to understand how ROS and redox probes work mechanistically [26,27]. Without this understanding the relative contributions of individual parameters to the response of cpYFP under physiological conditions have to be considered carefully in order to draw accurate conclusions.
carbonyl cyanide m-chlorophenylhydrazone
confocal laser-scanning microscopy
circularly permuted yellow fluorescent protein
electron transport chain
enhanced yellow fluorescent protein
green fluorescent protein
manganese superoxide dismutase
reactive oxygen species
yellow fluorescent protein
Markus Schwarzländer designed and performed the experiments and wrote the paper. Lee Sweetlove, David Logan and Mark Fricker supervised the research and co-wrote the paper.
We thank Dr Wang Wang (University of Washington, Seattle, WA, U.S.A.) and Professor Heping Cheng (Peking University, Bejing, China) for kindly providing the cpYFP construct, Dr Karl Morten (University of Oxford, Oxford, U.K.) for the gift of myxothiazol, Dr Daniel Kliebenstein (University of California, Davis, CA, U.S.A.) for the MnSOD antiserum, and Dr Ian Moore (University of Oxford, Oxford, U.K.) for the pOpOff2 vector. Stimulating discussions with Professor George Ratcliffe (University of Oxford, Oxford, U.K.) and Professor Andreas Meyer (University of Bonn, Bonn, Germany) have strengthened this work.
M.S. was funded by New College Oxford through a Weston Junior Research Fellowship.